Leveraging prior information to detect causal variants via multi-variant regression

Area of Science:

• Genetics and Bioinformatics
• Computational Biology
• Disease Association Studies

Background:

• Identifying causal variants for complex diseases is crucial for understanding disease mechanisms.
• Existing methods primarily focus on association, not causality, limiting the identification of true causal variants.
• Developing advanced statistical approaches is necessary to improve causal variant detection.

Purpose of the Study:

• To develop and evaluate a novel Bayesian hierarchical regression method for identifying causal sequence variants.
• To incorporate prior information on variant causality through effect weighting.
• To compare the proposed method against standard single-variant tests using simulations.

Main Methods:

• Developed a Bayesian hierarchical regression model incorporating prior probabilities of variant causality.
• Employed weighting schemes based on linkage disequilibrium with known Genome-Wide Association Study (GWAS) signals and sequence conservation (phastCons).
• Conducted simulation studies using both synthetic and real sequence variant data.

Main Results:

• The proposed Bayesian method demonstrated superior power in detecting causal variants compared to standard single-variant tests.
• Leveraging linkage disequilibrium and phastCons scores significantly enhanced the detection of causal variants.
• The method effectively controlled for false positives, improving the reliability of identified causal variants.

Conclusions:

• The developed Bayesian hierarchical regression method offers a powerful approach for identifying causal variants in complex diseases.
• Integrating linkage disequilibrium and sequence conservation data improves the accuracy and reduces false positives in causal variant discovery.
• This method advances the field of genetic association studies by focusing on causal variant identification.